Design and Development of Optical Coatings on Laser Bar Facets
D. Morton*a, I. Stevensona, M. Garciab
Denton Vacuum, LLC, 1259 N Church St., Moorestown, NJ, USA 08057;
b
Thales Research & Technology France, Domaine de Corbeville, 91404 Orsay cedex, France
a
ABSTRACT
There are a variety of optical coatings needed on laser bar facets to make them functional. There are also several
different types of laser bar facet materials to be coated which complicates the problem a little bit. These coatings
fall into three types; antireflection coatings, high reflectors and partial reflectors. A wide range of coating designs
and materials to be used in the coatings has been studied.
The anti reflection coating (AR) typically used is either a single layer coating, obviously consisting of one material,
or dual layer coatings consisting of two materials. There are a limited number of applications that may involve
larger number of layers if more than one wavelength or a wavelength band of more than 40 nm needs to be covered.
The single layer coatings usually do not provide very low reflectance. Dual layer coatings provide the ability to
design coatings with very low reflectance. Manufacturing process limitations allow for producing AR coatings with
a residual reflection in the 0.1% - 0.2% range. Although lower reflecting coatings can be designed, process control
parameters and optical measurement problems limit the coating manufacturer to this rang with R ≤ 0.2% being a
standard specification. This paper will discuss the various coating designs for achieving low reflectance on InP and
GaAs laser facets and the optical measurements problems.
High reflectors are relatively easy to design. There are a variety of material choices available. One choice
(Si/Al2O3) when applied for wavelengths below the absorption edge of silicon will result in a limit to how high the
reflectance can be made. The use of non-absorbing materials will allow the reflectance of the coatings to be
increased to the 99% range. The performance of various designs will be shown on different laser facet materials.
Partial reflecting mirrors 90/10 (R/T), 95/5 and other combinations can also be designed. These designs are used
where the smaller transmitted signal can be used as a signal for controlling the cooling of the laser facet. Although
these design are similar to the design of the high reflectors, using a symmetrical structure concept makes these
coating easier to design and to trim into performance requirements. These designs will also be presented and
discussed.
1. INTRODUCTION
As indicated in the abstract, we will be discussing the design of optical coatings on laser facets for a variety of
applications. The optical properties of the materials used in designing and the manufacturing process are typically
well known and published. However the process conditions used in various coating systems may vary as a result of
the construction of the system or deliberate variations in controlling deposition conditions. These differences can
result in variations from values as published. When designing these coatings it is critical that one uses the optical
film characteristics resulting from the coating system that will be used to fabricate the devices. Therefore it is
essential to characterize the optical properties of the films using deposition conditions that will be used in the
manufacturing process. Most organizations that make these devices have techniques to measure the optical
properties of films. One of the most popular techniques is VASE (variable angle spectroscopy) and information on
how to do this is available from the manufacturer of the equipment. Optical coating houses may or may not have
ellipsometers. However they will have spectrophotometers, which are used to measure the reflectance and
transmission of films as a function of the wavelength. This data can be used to extract the optical properties of the
films. This can be done by simple mathematics as published by Morton [1,2]. An alternative method would be to
use any one of the thin film design programs that are commercially available. These usually have a procedure for
entering measured data to extract the optical properties of optical film materials.
*dmorton@dentonvacuum.com; phone 903-498-4674; fax n/a; dentonvacuum.com
We will also show a typical configuration for an E-beam evaporator coating machine with ion assist that can be used
to coat the opposite faces of a laser facet in a single pump-down. We will then compare measured results for some
of the coating designs. One of the problems in verifying the spectral performance of antireflection coatings on laser
facet materials is the use of thin wafers as witness samples for spectral measurements. The cause of this problem
and the solution will be discussed.
2. DESIGNS
Single layer AR
The most common laser facet materials are InP and GaAs. These two materials have a fairly high refractive index
and can be perfectly antireflection coated with a low refractive index material equal to the square root of the
refractive index of the laser facet at that particular wavelength. Table I shows the refractive index of these materials
at various wavelengths, the refractive index for the perfect antireflection coating and the residual reflectance
expected using ion-assisted quarter wave optical thick Al2O3 for the single layer AR. Al2O3 was chosen for the AR
since it is the most common one used by manufacturers. GaSb is a material which we expect to see being used for
devices active in the 2 mm – 5 mm spectral region. The refractive index is about 3.8 in those regions and a single
layer AR would have a residual reflectance of about 0.9%
Table I
λ (nm)
N of InP [3]
√n
R (%)
N of GaAs [4]
√n
R (%)
810
~3.40
1.84
0.94
3.66
1.94
2.09
980
3.346
1.829
0.956
3.522
1.91
1.59
1310
3.205
1.789
0.626
3.423
1.887
1.25
1550
3.1665
1.78
0.54
3.423
1.884
1.26
Materials such as SiO, MgO and Y2O3 would result in lower residual reflection since they would be a better
refractive index match. However these film materials probably would not be as durable. Al2O3 is usually the
material of choice because it is very stable material and resists changes in structure or composition while being
subjected to further device assembly steps.
Two layer AR
Lower reflectance can be achieved by going to two layers. The new material would have a higher refractive index
so that the combination of the two materials would result in exactly zero reflectance. These thickness values are
normally found by using a function in the thin film software that varies the thickness of each material to minimize a
merit function (normally the actual value minus the target value over a wavelength range raised to some power). As
an example, we have done this using TiO2 as the additional material. Thus the design becomes GaAs (or
InP)/Al2O3/ TiO2 /air. The optical properties of the Al2O3 will be the same as for the single layer results shown in
Table I. Deposition conditions for the TiO2 will be those to achieve a moisture stable film (n=2.48 @ 560 nm). The
thickness (in nanometers) of the layers needed to achieve 0% reflectance for InP and GaAs at the previously selected
wavelengths is recorded in Table II. Although TiO2 was chosen for this example, other high index materials would
work equally well. However the film thickness would have to be varied due to the different optical properties.
Table II
λ (nm)
n of InP [3] Al2O3 (nm) TiO2 (nm)
810
980
~3.40
3.346
1310
1550
3.205
3.165
69.2
23.5
79
123.7
132.8
32.6
33.3
47.3
n of GaAs [4]
3.66
3.522
3.423
3.423
Al2O3 (nm) TiO2 (nm)
57.1
31.3
77.3
104.9
124.1
33.5
44.3
52.3
Partial AR Coatings
There are some applications where a controlled level for the reflectance of the AR is desired, usually to optimise the
resonance in the laser. This can be done by displacing the quarter wave optical thickness to a thinner or higher value
if the reflectance desired is greater than that achieved with a single layer antireflection coating. As an example
consider the desire to have a 4% reflecting surface for GaAs. The single layer centred AR would be 200.4 nm thick
Al2O3 and result in a reflection of 1.25%. Using a thickness of 168 nm would displace the AR to a shorter
wavelength resulting in a 4% reflectance at 1310 nm. Using a thickness of 232.9 nm would displace the AR to a
longer wavelength resulting in a 4% reflectance at 1310 nm. See Figure 1 for a plot of the reflectance of these three
films as a function of wavelength. The problem of using this method is that the manufacturing tolerance is reduced
for the displaced coatings.
Figure 1. Reflectance vs. wavelength for different Al2O3 as indicated.
One can also displace the central wavelength of a two material two layer coating to get a higher reflectance than that
achieved with the “perfect” AR coating. However, this also results in a reduced manufacturing tolerance for the
displaced coating. The most logical optical way to do this is to design a symmetrical 3-layer coating consisting of
two materials that results in the required reflectance. This structure is the easiest to adjust when doing the original
development of the coating and will have a better manufacturing tolerance (see Figure 2).
Figure 2. Reflectance vs. wavelength for symmetrical period as indicated (layer thickness is in nm).
High Reflector Mirrors
High reflector coatings are merely stacks of high and low index of materials with sufficient number of layers to give
a high enough reflectance. The refractive index ratio of the two chosen materials will determine the number of
layers needed. Fewer layers will be needed as the refractive index ratio becomes greater. For this reason many
manufacturers chose to use a combination of silicon and aluminium oxide for the mirror stacks. Typically 6 to 8
layers will be sufficient. The problem with using this combination below 1 micron is that the silicon becomes
absorbing and limits the ultimate reflectance to about 95% -96% at 810 nm. Using non-absorbing materials such as
TiO2/SiO2 or TiO2/Al2O3 can result in a reflectance as high as 99% or greater, depending on deposition conditions.
The theoretical reflection of two coating combinations is compared in Figure 3. The higher index ratio Si/Al2O3
requires fewer layers (only 6) and has a much broader high reflection region than the lower index ratio TiO2/Al2O3
(requiring 14 layers). The theoretical refractive index of silicon films is never achieved in standard E-beam
evaporation where there are alternating layers of silicon and an oxide. This is due to the fact that residual oxygen
left over from the oxide deposition combines with the silicon producing a lower index partially oxidized silicon film.
Therefore the actual deposited stack will have slightly lower reflectance and not quite as broad a high reflection
region. This is not a serious problem since in the worst an additional pair might be needed to bring the reflection up.
Figure 3. Comparison of high reflector stacks using Al2O3 as the low index material and Si or TiO2 as the high index
material.
Partial High Reflectors
There are situations where there is a need for some of the incident energy needs to be transmitted through the mirror
(again to optimise the resonance in the laser). This is easy to do since the peak reflection is dependent on the
number of layer pairs making up the mirror. For example, consider a mirror coating design of (L H)x. The H stands
for a QWOT (quarter wave optical thickness of a high index material and the L stands for a QWOT of a low index
material. The lowercase x represents the number of times the LH pair is repeated in the mirror design. If x is 3 then
there are a total of 6 layers in the design and if x is 5 there are 10 layers in the design. The following plot (Figure 4)
shows the development of a mirror stack on GaAs as the x value is increased from 3 to 9 using TiO2 and Al2O3 as
the coating materials. If the natural reflectance of a given number of pairs is what is needed, use it. Typically a
desired reflectance and tolerance is not achieved with just QWOT layers. In that case one would use a design that
gives a reflectance greater than the desired value and then tune the design in by symmetrically varying the thickness
of the last three layers. As an example for this consider a design of (L H)x where x is 5 and the high and low index
layers are TiO2 and Al2O3 respectively.
Figure 4. Reflectance plots of x = 3, 5, 7 and 9 layer pairs using TiO2 for the high index material and Al2O3 as the low
index material.
For example, if 80% is the desired reflectance, use the 3 pair (or 6 layer) design, vary the thickness of layers 4, 5 and
6 (while keeping layers 4 and 6 the same thickness) until you have the desired reflectance (as shown in Figure 5).
Figure 5. Comparing GaAs/LHLHLH design to a GaAs/LHL .648H 1.62L .648H design. The multiplier applies to the
normal QWOT thickness.
It is important to note that all of the partial reflector designs have rounded tops and therefore the desired reflectance
level is only achieved over a narrow wavelength band, requiring good thickness control when depositing these
coatings. The lower the reflectance value the narrower the band. Whereas high reflectance levels are achievable
over wider reflectance bands as the number of pairs are increased and the reflectance approaches 100%. Layer
thickness control for coatings with very high reflectors are not nearly as critical as for AR coatings.
3. A Typical Flip Evaporator
Any high vacuum evaporation chamber can be used to do laser facet coatings. However, a typical size,
configuration with specific components has evolved over time. Since laser facets are relatively small. They are
usually coated in a long bar configuration and then sliced to size after coating. Also the chamber size needs to be
only as big as is needed to accommodate the equipment inside the chamber and the tooling to hold the parts.
Electron beam evaporation with an ion beam assist is the preferred deposition method. Typical tooling configuration
is to have a single rotation fixture that will hold up to 8-three inch jigs which can be flipped to coat both sides in a
single pump-down. The parts will be located on a relatively small area on the radius of the tool and will result fairly
good deposition uniformity. However a mask is sometimes included in order to achieve the optimum uniformity.
Adequate thickness control is normally achieved with a quartz crystal monitor. Further improvements in thickness
control can be achieved by adding an optical monitor. Meeting these requirements usually results in a coating
chamber with a twenty-six inch diameter and a height of about thirty-two inches. Larger chambers can be used but
are not required. A typical system is shown in Figure 6.
a.
b.
Figure 6. a) Internal view of a laser bar flip coater. b) View of flip tooling.
4. Results from a Flip Evaporator
All of the coating design types discussed in the design portion of this paper have been fabricated in coating
equipment in our facility. Most of the coatings were done in equipment being built specifically for
telecommunication manufacturers. Some of the coatings were done in our application laboratory as a development
project. A few coatings have been done in our coating division as a toll coating service for customers who need
production prior to getting there coating equipment built, installed and commissioned for production. Although we
used GaAs laser facets in the design portion of this paper, all of the measured results were for either InP or GaAs
facets. The differences in the designs for the two different facet materials is slight adjustments in film thickness
needed to accommodate refractive index differences between the InP and the GaAs.
Evaluations of coatings done in our facility were done using a spectrophotometer to compare spectral results to
design criteria. Evaluation of any actual devices was done by the customer buying equipment and the only feedback
which we had for their evaluation was changes which might be desired or that they were satisfied with the device
performance. We did not receive back any information on specific performance results. The usual substrate used
for spectral evaluation was a wafer of the same (or similar) material as for the devices being fabricated. These
wafers had one surface polished and the second surface etched. This configuration has a potential problem in that
the quality of the back surface etch on samples supplied to us were not consistent. Accurate reflectance
measurements on our spectrophotometer require that there should be no spectral reflectance coming back from the
second surface of the witness samples. Witness samples for most optical work requires the reflectance samples have
the back surface polished at a wedge so that any back surface reflectance is blocked out of the beam by an aperture.
An alternative method is to roughen the back surface by grinding it to such a rough texture that all the light
reflecting from the back surface is scattered form the beam. Initial measurements during our development work for
AR coating resulted in inconsistencies that we could not understand. We felt that the second surface might be a
problem and therefore scanned the second surface reflectance of InP wafers from several sources. The results
showed significant variations as shown in Figure 7.
R (%)
InP Etched Surface R
7
6
5
4
3
2
1
0
1000
1100
1200
1300
1400
1500
1600
1700
Wavelength (nm)
Figure 7. Spectrophotometer measured reflectance of etch surface of various InP wafers.
When the front polished surfaces of these wafers were coated with the same AR coating, the results were different.
The measured reflectance was higher of the samples with the higher back surface reflectance than those with the
lower reflectance. Furthermore, we found that by painting the back surface with a flat black paint (a technique often
used in optical coating manufacturing), the reflectance could be reduced even more. Upon further investigation we
found that the company which supplied the lower reflecting back surface samples had recognized this problem and
were grinding the back surface of their wafers with a 1 micron grit powder to give them consistent results.
As a result of this study, our coating department requires customers to supply samples with a back surface
reflectance of less than 2% (≤1% being preferred). We know that there is still an undefined residual reflectance but
these samples our good enough to guarantee us to consistently confirm reflectance of ≤ 0.2%, and we typically get
about 0.1%.
Figure 8 shows the computed “perfect” AR on InP at 1550 nm where the coating materials are TiO2 and SiO2.
Figure 8. Close to the “perfect” AR on InP using TiO2 and
Al2O3 films.
Figure 9. A comparison of a 1550 nm AR to the original
design deposited on InP. Also showing recent improved
performance of dual band AR.
The measured spectral results of early depositions on InP are shown in figure 9. As can be seen the measured curve
is not as low as the design due in part to random errors during deposition but mainly because there is residual
reflectance from the back surface as discussed previously. In practice the design can be improved by doing a series
of depositions and making slight variation in layer thickness to optimise empirically previous calibrations.
In the field of laser signal processing and telecommunications, the term Optical Return Loss (ORL) is used to
describe the amount of light reflected at the end of a fiber or a laser bar facet. It is measured in decibels (dB) and is a
logarithmic expression of the ratio of returned to incident light.
ORL = 10log10 [Returned/Incident] (in dB)
It is desirable to reduce the return loss to the smallest possible value, and AR coatings are used extensively to
accomplish this. The returned signal may not be exactly equal to the Fresnel loss at the coated interface: scattering,
polish angle etc can cause a discrepancy, but if we ignore these effects it is simple to construct a table of reflectance
and ORL values.
Table III
ORL (dB)
-14.5
-20
-27
-30
-35
-40
-45
-50
Equivalent Reflectance (%)
~3.5
1.0
0.2
0.1
0.03
0.01
0.003
0.001
Uncoated silica
Typical high precision AR
Ultra Low Loss AR
In our previous discussion, we stated that a typical high performance AR coating could have a residual reflectance of
0.1% – 0.2%, or –27 to –30dB. For telecommunications applications in recent years, an ORL of –30dB has been
regarded by some as a barely acceptable minimum, with the majority of industry specifications requiring –35 to –
40dB, some even more. These correspond to reflectance values an order of magnitude smaller than had, until
recently, been the industry standard. To achieve such low reflectance values the coating process must be capable of
extremely precise control of both film thickness and film index. The requirements have forced us to improve not
only our manufacturing capability but also our measurement technology to the point where we can consistently
produce such ultra low loss coatings and prove that they meet specification.
Figure 9 shows the measured reflectance spectrum of a dual band low loss AR coating. Fibers with this coating were
measured with ORL values of –38dB at the primary wavelength (1310nm) and –32dB at the secondary wavelength
(1550nm). These are quite consistent with the measured reflectance values of 0.016% and 0.06% respectively.
High reflector mirrors are less difficult to make. The reflectance bands are quite broad and they are much less
susceptible to random errors in film thickness. Early work in our facility was for a customer who required Si/Al2O3
stacks for the high reflector on InP and the wavelengths of interest was 980 nm. Initial theoretical design
performance and actual performance for an eight-layer design InP/LHLHLHLH is shown in figure 10. The optical
properties of the Al2O3 were measured. The optical properties of the Si used were those supplied with the Essential
Macleod software for Si (FILM) since we had no way for us to measure the properties of a strongly absorbing film.
The film was actually deposited on a glass substrate since at the time we did not have InP wafers available.
Theoretical calculations for the same stack on InP and glass showed little difference in the high reflectance regions.
Therefore the difference shown here are the fact that the silicon film properties used in the theoretical model are not
accurate. The center of the designed coating is longer than the center of the actual deposited coating. This is due to
the fact that the initial single layer coatings are always close but not as close as desired. The calibrations can be
adjusted for future coatings so that there is a better correlation.
There is a major variation from the peak reflectance values due to the actual and assumed optical properties of the
silicon due to the fact that we had no reliable way to determine the actual optical properties of the silicon in the
region where it was highly absorbing (below 1000 nm). An alternative mirror design would be to use TiO2 in place
of the silicon. TiO2 is non-absorbing in this spectral region. However the refractive index is lower and therefore
more layers are required. A comparison of a 12-layer design for TiO2/Al2O3 (LHLHLHLHLHLH) and the coating
deposited on a GaAs wafer is shown in figure 11.
Figure 10. A comparison of a theoretical (R-calc.) LHLHLHLH Al2O3/Si reflector on glass and the first deposited stack
on glass substrates.
Designs of either of these material combinations can easily be shifted out to 1310 nm or 1550 nm. Out at the longer
wavelengths, the comparison between design and actual results for the Si/Al2O3 combinations are closer since we
can get better optical properties for the silicon. However, silicon is a highly reactive material therefore it is critical
that the deposition process is not started until the vacuum level is very low (typically 2x10-6 Torr) and that the
amount of gas entering the chamber either through leaks or deliberate backfills are kept low so as to provide little or
no oxygen for the silicon to combine with. This will result in silicon films with higher refractive index and mirror
stacks with maximum width.
Figure 11. A comparison of a theoretical (R-design) LHLHLHLHLHLH Al2O3/TiO2 reflector on GaAs and the first deposited
stack on a GaAs wafer.
We had previously discussed partial reflectors where the coating design had the last three layers with thickness
varied symmetrically so that the partial reflectance value equalled a required value. We have never made this
coating although it is the preferred design method to do so. We have used an alternative technique for one customer
who desired the coating to have a peak reflectance of 90% ± 2% on InP at 1550 nm. The materials of choice were
silicon and Al2O3. A four-layer design LHLH has a reflectance of 95% and the 5-layer design of LHLHL would be
about 87%. Once the first 4 layers have been deposited the peak reflectance value will start shifting down and to a
longer wavelength as the 5th lower index layer is deposited. At some thickness for the 5th layer the peak wavelength
will reach a maximum value and start shifting back while the peak reflectance value continues to decrease to the
minimum value for the final thickness of the L layer. At some intermediate thickness value for the L layer the
reflectance will be the required value. An example of this is shown in figure 12.
Figure 12. Plots showings theoretical technique for designing partial reflector with only outer layer thickness being varied.
In order to actually make this design the desired central peak wavelength must be divided by the central wavelength
peak of the altered design and this scaling factor applied to the thickness of all the layers during deposition. When
this is done the results of the first run done for this project are shown in figure 13. Again the final results are not
exactly as would be expected, however, in this case they are within specification. The actual deposition results can
be improved by making the outer layer slightly thicker will both move it on wavelength and increase the reflectance
slightly. If desired, the layer ratios can also be altered slightly to further improve the results.
Figure 13. Comparison of 90% reflector design on InP and initial results.
CONCLUSIONS
In this paper we have shown a variety of designs and thin film material combinations for antireflection coatings and
high reflectors for laser bar facets. We have also discussed some of the problems which optical coating
manufacturers will have in measuring the performance of antireflection coatings on wafers that are normally
supplied or used as reference samples. And finally we have shown examples of coatings made in various equipment
in our coating facility and compared them to the original designs.
REFERENCES
1) D. E. Morton, “Characterizing Optical Thin Films (I)”, Vacuum Technology & Coating, September 2001, pp.2431.
2) D. E. Morton, “Characterizing Optical Thin Films (II)”, Vacuum Technology & Coating, November 2001, pp.4248.
3) P. Klocek ed., “Handbook of Infrared Optical Materials”, Marcel Dekker, Inc. New York, (1991) p 275.
4) Extracted from measured spectral data.